Hierarchical Conformal Prediction for Clustered Data with Missing Responses
Estimate conditional density pi(y|x) via quantile process + quotient e...
Fit missingness propensity model P(delta=1 | X) from pooled data
Simulate clustered continuous outcomes with covariate-dependent MAR mi...
HCP conformal prediction region with repeated subsampling and repeated...
HCP prediction wrapper for multiple measurements with optional per-pat...
Plot HCP prediction intervals (band vs covariate or intervals by patie...
Implements hierarchical conformal prediction for clustered data with missing responses. The method uses repeated cluster-level splitting and within-cluster subsampling to accommodate dependence, and inverse-probability weighting to correct distribution shift induced by missingness. Conditional densities are estimated by inverting fitted conditional quantiles (linear quantile regression or quantile regression forests), and p-values are aggregated across resampling and splitting steps using the Cauchy combination test.